Experimental verification of environmental models for adaptive detection and estimation in HF skywave radar

Performance of adaptive detection algorithms such as GLRT, AMF, and ACE are usually evaluated assuming training data consisting of i.i.d. Gaussian noise samples. While this assumption is often true, radar signal processing and spatial over-sampling in the array may introduce correlation between data samples spatially and temporally. Even in cases where the external noise is i.i.d. Gaussian, this sensor induced correlation can significantly change the convergence rate, discrimination capability, and false alarm rate of adaptive detectors. Receiver per element measurements from a large HF radar array are used to confirm these effects and the potential performance impact on hypothesis testing performance is examined.